Decisions · Policy · United States

The globalization of copyright exceptions for AI training

Matthew Sag (Emory University) and Peter K. Yu (Texas A&M University) · 2025 · Proposal · non-binding

74 Emory Law Journal 1163 (2025)

AI training and copyright

Can AI be trained on protected works?

No position

Jurisdictions are converging on allowing non-expressive AI training in some, but not all, situations

Why hereNot placed: a comparative survey of how jurisdictions treat AI training. It reports an emerging equilibrium rather than recommending one.

What it says

Surveying the US, Israel, Japan, the UK, the EU, Singapore, China and the UAE, the authors find an emerging international equilibrium in which AI training may be permitted without authorisation in some situations, driven by the idea and expression distinction, AI competition and a race to the middle in reforms.

A comparative map showing most systems allow some training without permission.

Sources

This record

Data as of 2026-10-21.

Drafted with AI assistance and checked against the sources linked above; not yet reviewed by a lawyer. Methodology

Changes

  • 2026-10-20Reclassified as descriptive work: no longer coded as a position (was in the middle) or counted as backing the proposal that training should generally be allowed.